The experimental data on adsorption of Fe and CN of a ferrocyanide complex onto raw and acid-activated clinoptilolite/sepiolite on the basis of detention time and particle size was modelled by a linear and a non-linear approach. The linearized best-fit isotherm selection method and non-linear error minimization was applied through Freundlich, Langmuir and Temkin isotherms. ERRSQ, MPSD, HYBRID and ARE error functions were minimized by a developed MATLAB script to determine the isotherm parameters in non-linear optimization. The complex was not adsorbed as whole anions but the Fe and CN were adsorbed separately. 0.65 mg Fe/L.min and 4.84 mg CN/L.min initial adsorption rates were achieved with acid activated clinoptilolite. The Fe adsorption was not as successful as CN. The adsorption of Fe and CN was described by Freundlich and Langmuir isotherms respectively. The differences between the predicted isotherm parameter sets of linear models and minimized error function models indicated that both the best-fit isotherm selection and the isotherm constant determinations can be performed properly by error minimization as well as by conventional linear best fit modelling approach.